Update configuration_hf_nomic_bert.py
Browse files
configuration_hf_nomic_bert.py
CHANGED
|
@@ -4,7 +4,8 @@ from transformers import GPT2Config
|
|
| 4 |
class NomicBertConfig(GPT2Config):
|
| 5 |
model_type = "nomic_bert"
|
| 6 |
|
| 7 |
-
def __init__(
|
|
|
|
| 8 |
prenorm=False,
|
| 9 |
parallel_block=False,
|
| 10 |
parallel_block_tied_norm=False,
|
|
@@ -26,6 +27,7 @@ class NomicBertConfig(GPT2Config):
|
|
| 26 |
pad_vocab_size_multiple=1,
|
| 27 |
tie_word_embeddings=True,
|
| 28 |
rotary_scaling_factor=1.0,
|
|
|
|
| 29 |
**kwargs,
|
| 30 |
):
|
| 31 |
self.prenorm = prenorm
|
|
@@ -49,5 +51,6 @@ class NomicBertConfig(GPT2Config):
|
|
| 49 |
self.dense_seq_output = dense_seq_output
|
| 50 |
self.pad_vocab_size_multiple = pad_vocab_size_multiple
|
| 51 |
self.rotary_scaling_factor = rotary_scaling_factor
|
|
|
|
| 52 |
|
| 53 |
-
super().__init__(**kwargs)
|
|
|
|
| 4 |
class NomicBertConfig(GPT2Config):
|
| 5 |
model_type = "nomic_bert"
|
| 6 |
|
| 7 |
+
def __init__(
|
| 8 |
+
self,
|
| 9 |
prenorm=False,
|
| 10 |
parallel_block=False,
|
| 11 |
parallel_block_tied_norm=False,
|
|
|
|
| 27 |
pad_vocab_size_multiple=1,
|
| 28 |
tie_word_embeddings=True,
|
| 29 |
rotary_scaling_factor=1.0,
|
| 30 |
+
max_trained_positions=2048,
|
| 31 |
**kwargs,
|
| 32 |
):
|
| 33 |
self.prenorm = prenorm
|
|
|
|
| 51 |
self.dense_seq_output = dense_seq_output
|
| 52 |
self.pad_vocab_size_multiple = pad_vocab_size_multiple
|
| 53 |
self.rotary_scaling_factor = rotary_scaling_factor
|
| 54 |
+
self.max_trained_positions = max_trained_positions
|
| 55 |
|
| 56 |
+
super().__init__(**kwargs)
|